Plasma p-tau217 explained 16% of cognitive timing in the community and 64% in a trial cohort

Design

cohort study with non-linear mixed-effects modelling of biomarker and cognitive trajectories

Population

2369 population-based ageing study participants and 1591 from an Alzheimer's-enriched cohort

Primary outcome

association between individual timing of biomarker progression and timing of verbal learning decline

Effect

plasma p-tau217 explained 16% of timing variability in the population sample, 64% in the enriched cohort

Non-linear mixed-effects models were fitted across 2369 participants of a population-based ageing study and 1591 from an Alzheimer's disease neuroimaging cohort, estimating for each person how much earlier or later than average they progressed on plasma phosphorylated tau 217, amyloid PET, tau PET and a verbal learning test. The question was whether the timing of biomarker change predicts the timing of cognitive change in the same individual.

It does, but how strongly depends entirely on who is in the sample. Correlations between biomarker and cognitive timing were moderate in the population cohort (R = 0.38 to 0.47) and much stronger in the enriched cohort (R = 0.74 to 0.81). Plasma p-tau217 accounted for 16% of the variability in the timing of verbal learning decline in the population sample against 64% in the enriched one. APOE ε4 carriage was associated with earlier progression on the biomarkers.

The expected ordering - amyloid, then tau, then cognition - held up to about age 90, after which cognition was estimated to become abnormal before the tau markers, most plausibly because non-Alzheimer's co-pathology is doing the damage in the oldest old. The practical reading is that plasma p-tau217 performs about as well as amyloid PET as a progression marker, and that performance figures quoted from enriched research cohorts will not be reproduced in a memory clinic.

• Discount biomarker performance figures derived from enriched research cohorts when applying them to unselected clinic patients.
• Plasma p-tau217 tracked cognitive timing about as well as amyloid PET - relevant where PET is unavailable or unaffordable.
• In patients over 90, do not assume a normal tau marker excludes a neurodegenerative cause of decline.
• Use APOE ε4 status as a marker of earlier progression, not as a diagnostic test.
• A biomarker that explains 16% of timing variability is not a prognosis for an individual.

Why it matters

It gives a number to the gap between biomarker performance in research cohorts and in the clinic, which is where these tests are now being sold.

Don't overread it

This is modelling of observational cohort data, and the biomarker timing estimates depend on the model's assumptions about the shape of progression.

The statistics, in plain English

The 16% versus 64% difference is not a contradiction: an enriched cohort is deliberately built from people who have the disease and will progress, which inflates every correlation between a disease marker and an outcome. A correlation of 0.38 to 0.47 in a community sample means the biomarker and the cognitive change move together loosely - useful at group level, weak for telling one patient when their decline will begin.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.